gsd-pi 2.73.1 → 2.74.0-dev.0306a2e

package/src/resources/extensions/gsd/templates/PREFERENCES.md

@@ -39,6 +39,24 @@ dynamic_routing:
   budget_pressure:
   cross_provider:
   hooks:
+uok:
+  enabled: true
+  legacy_fallback:
+    enabled: false
+  gates:
+    enabled: false
+  model_policy:
+    enabled: false
+  execution_graph:
+    enabled: false
+  gitops:
+    enabled: false
+    turn_action: status-only
+    turn_push: false
+  audit_unified:
+    enabled: false
+  plan_v2:
+    enabled: false
 auto_visualize:
 auto_report:
 parallel:

package/src/resources/extensions/gsd/tests/auto-loop.test.ts

@@ -688,8 +688,8 @@ test("autoLoop exits on terminal blocked state", async (t) => {
 
   assert.ok(deps.callLog.includes("deriveState"), "should have derived state");
   assert.ok(
-    deps.callLog.includes("stopAuto"),
-    "should have called stopAuto for blocked state",
+    deps.callLog.includes("pauseAuto"),
+    "should have called pauseAuto for blocked state",
   );
   assert.ok(
     !deps.callLog.includes("resolveDispatch"),
@@ -1267,7 +1267,7 @@ test("auto-loop.ts barrel re-exports autoLoop, runUnit, and resolveAgentEnd", ()
   );
 });
 
-test("auto.ts startAuto calls autoLoop (not dispatchNextUnit as first dispatch)", () => {
+test("auto.ts startAuto dispatches through the UOK kernel wrapper (legacy loop adapter)", () => {
   const src = readFileSync(
     resolve(import.meta.dirname, "..", "auto.ts"),
     "utf-8",
@@ -1279,8 +1279,12 @@ test("auto.ts startAuto calls autoLoop (not dispatchNextUnit as first dispatch)"
   const fnBlock =
     fnEnd > -1 ? src.slice(fnIdx, fnEnd) : src.slice(fnIdx, fnIdx + 5000);
   assert.ok(
-    fnBlock.includes("autoLoop("),
-    "startAuto must call autoLoop() instead of dispatchNextUnit()",
+    fnBlock.includes("runAutoLoopWithUok("),
+    "startAuto must dispatch through runAutoLoopWithUok()",
+  );
+  assert.ok(
+    fnBlock.includes("runLegacyLoop: autoLoop"),
+    "startAuto must preserve the legacy autoLoop adapter in kernel dispatch",
   );
 });
 

package/src/resources/extensions/gsd/tests/auto-model-selection.test.ts

@@ -227,6 +227,26 @@ test("model change notify in selectAndApplyModel is gated behind verbose flag",
   );
 });
 
+test("model policy resolves candidates from the policy-eligible pool", () => {
+  const src = readFileSync(join(__dirname, "..", "auto-model-selection.ts"), "utf-8");
+  assert.ok(
+    src.includes("const resolutionPool = uokFlags.modelPolicy ? routingEligibleModels : availableModels"),
+    "selectAndApplyModel should resolve model IDs against policy-eligible models when model policy is enabled",
+  );
+});
+
+test("model policy receives task metadata for requirement-vector decisions", () => {
+  const src = readFileSync(join(__dirname, "..", "auto-model-selection.ts"), "utf-8");
+  assert.ok(
+    src.includes("taskMetadata: taskMetadataForPolicy"),
+    "applyModelPolicyFilter should receive task metadata so requirement vectors are unit-aware",
+  );
+  assert.ok(
+    src.includes("extractTaskMetadata(unitId, basePath)"),
+    "execute-task dispatch should derive metadata before policy filtering",
+  );
+});
+
 test("resolveModelId: anthropic wins over claude-code when session provider is not claude-code", () => {
   const availableModels = [
     { id: "claude-sonnet-4-6", provider: "claude-code" },

package/src/resources/extensions/gsd/tests/auto-post-unit-step-message.test.ts

@@ -0,0 +1,53 @@
+// GSD-2 — Tests for step-mode completion messages in auto-post-unit
+
+import test from "node:test";
+import assert from "node:assert/strict";
+
+import { buildStepCompleteMessage, STEP_COMPLETE_FALLBACK_MESSAGE } from "../auto-post-unit.ts";
+import type { GSDState } from "../types.ts";
+
+function makeState(overrides: Partial<GSDState>): GSDState {
+  return {
+    activeMilestone: null,
+    activeSlice: null,
+    activeTask: null,
+    phase: "executing",
+    recentDecisions: [],
+    blockers: [],
+    nextAction: "",
+    registry: [],
+    ...overrides,
+  };
+}
+
+test("buildStepCompleteMessage: milestone complete surfaces review guidance", () => {
+  const msg = buildStepCompleteMessage(makeState({ phase: "complete" }));
+  assert.match(msg, /milestone finished/);
+  assert.match(msg, /\/gsd status/);
+  assert.doesNotMatch(msg, /Next:/);
+});
+
+test("buildStepCompleteMessage: mid-flight step includes next unit label and /clear hint", () => {
+  const state = makeState({
+    phase: "executing",
+    activeSlice: { id: "S01", title: "Core" },
+    activeTask: { id: "T03", title: "Wire notify" },
+  });
+  const msg = buildStepCompleteMessage(state);
+  assert.match(msg, /Next: Execute T03: Wire notify/);
+  assert.match(msg, /\/clear/);
+  assert.match(msg, /\/gsd to continue/);
+});
+
+test("buildStepCompleteMessage: unknown phase falls back to generic continue label", () => {
+  // Cast to bypass Phase union so we exercise the default branch of describeNextUnit.
+  const state = makeState({ phase: "totally-unknown" as unknown as GSDState["phase"] });
+  const msg = buildStepCompleteMessage(state);
+  assert.match(msg, /Next: Continue/);
+  assert.match(msg, /\/clear/);
+});
+
+test("STEP_COMPLETE_FALLBACK_MESSAGE: used when deriveState throws, still points users at /clear + /gsd", () => {
+  assert.match(STEP_COMPLETE_FALLBACK_MESSAGE, /\/clear/);
+  assert.match(STEP_COMPLETE_FALLBACK_MESSAGE, /\/gsd/);
+});

package/src/resources/extensions/gsd/tests/auto-project-root-env.test.ts

@@ -13,11 +13,15 @@ test("auto-mode captures GSD_PROJECT_ROOT before entering the dispatch loop", ()
   const resumeCallIdx = source.indexOf("captureProjectRootEnv(s.originalBasePath || s.basePath);");
   assert.ok(resumeCallIdx > -1, "auto.ts should capture GSD_PROJECT_ROOT before resume autoLoop");
 
-  const firstAutoLoopIdx = source.indexOf("await autoLoop(ctx, pi, s, buildLoopDeps());");
-  assert.ok(firstAutoLoopIdx > -1, "auto.ts should invoke autoLoop()");
+  const firstLoopIdxCandidates = [
+    source.indexOf("await runAutoLoopWithUok({"),
+    source.indexOf("await autoLoop(ctx, pi, s, buildLoopDeps());"),
+  ].filter((idx) => idx > -1);
+  const firstAutoLoopIdx = firstLoopIdxCandidates.length > 0 ? Math.min(...firstLoopIdxCandidates) : -1;
+  assert.ok(firstAutoLoopIdx > -1, "auto.ts should invoke the auto dispatch loop");
   assert.ok(
     resumeCallIdx < firstAutoLoopIdx,
-    "auto.ts must set GSD_PROJECT_ROOT before the first autoLoop() call",
+    "auto.ts must set GSD_PROJECT_ROOT before the first loop call",
   );
 });
 

package/src/resources/extensions/gsd/tests/auto-start-model-capture.test.ts

@@ -33,8 +33,12 @@ test("bootstrapAutoSession checks manual session override before preferences", (
   assert.ok(manualIdx > -1, "auto-start.ts should read session model override first");
 
   // resolveDefaultSessionModel() should still be called for fallback behavior
-  const preferredIdx = source.indexOf("const preferredModel = resolveDefaultSessionModel(");
-  assert.ok(preferredIdx > -1, "auto-start.ts should call resolveDefaultSessionModel()");
+  const preferredIdx = source.indexOf("const preferredModel = ");
+  assert.ok(preferredIdx > -1, "auto-start.ts should build preferredModel");
+  assert.ok(
+    source.indexOf("resolveDefaultSessionModel(") > -1,
+    "auto-start.ts should call resolveDefaultSessionModel()",
+  );
 
   // Session provider should be passed for bare model ID resolution
   const withProviderIdx = source.indexOf("resolveDefaultSessionModel(ctx.model?.provider)");
@@ -47,6 +51,51 @@ test("bootstrapAutoSession checks manual session override before preferences", (
     manualIdx < snapshotIdx && preferredIdx < snapshotIdx,
     "manual override and preference fallback must be resolved before building startModelSnapshot",
   );
+
+  // The validated preferred model must still appear as one of the snapshot
+  // sources so PREFERENCES.md continues to win over a stale settings.json
+  // default for built-in providers.
+  const snapshotBlock = source.slice(snapshotIdx, snapshotIdx + 400);
+  assert.ok(
+    snapshotBlock.includes("validatedPreferredModel") || snapshotBlock.includes("preferredModel"),
+    "startModelSnapshot must still consider preferredModel for built-in providers",
+  );
+});
+
+test("bootstrapAutoSession prefers session model over PREFERENCES.md when provider is custom (#4122)", () => {
+  // Custom providers (Ollama, vLLM, OpenAI-compatible proxies) live in
+  // ~/.gsd/agent/models.json, not PREFERENCES.md.  When the user picks one
+  // via /gsd model, that selection must win over any preferredModel from
+  // PREFERENCES.md, otherwise auto-mode tries to start a built-in provider
+  // the user is not logged into and pauses with "Not logged in".
+  const customCheckIdx = source.indexOf("isCustomProvider(ctx.model?.provider)");
+  assert.ok(
+    customCheckIdx > -1,
+    "auto-start.ts should call isCustomProvider() to detect custom-model sessions",
+  );
+
+  // sessionProviderIsCustom must gate preferredModel resolution so that when the
+  // session provider is custom, preferredModel is null and PREFERENCES.md is
+  // skipped entirely — the snapshot then falls through to ctx.model.
+  const gateIdx = source.indexOf("sessionProviderIsCustom");
+  assert.ok(gateIdx > -1, "auto-start.ts should bind sessionProviderIsCustom");
+
+  const preferredIdx = source.indexOf("const preferredModel = ");
+  assert.ok(preferredIdx > -1, "auto-start.ts should build preferredModel");
+
+  const preferredBlock = source.slice(preferredIdx, preferredIdx + 200);
+  assert.ok(
+    preferredBlock.includes("sessionProviderIsCustom"),
+    "preferredModel must be gated on sessionProviderIsCustom so PREFERENCES.md is skipped for custom providers",
+  );
+
+  const snapshotIdx = source.indexOf("const startModelSnapshot = ");
+  assert.ok(snapshotIdx > -1, "auto-start.ts should build startModelSnapshot");
+
+  assert.ok(
+    customCheckIdx < preferredIdx && preferredIdx < snapshotIdx,
+    "isCustomProvider() must be evaluated before preferredModel, which must be resolved before startModelSnapshot",
+  );
 });
 
 test("bootstrapAutoSession validates preferred model against live registry auth (#unconfigured-models)", () => {

package/src/resources/extensions/gsd/tests/cold-resume-db-reopen.test.ts

@@ -25,8 +25,12 @@ console.log("\n=== resume path refreshes resources and opens DB before rebuildSt
 const resumeSectionStart = autoSrc.indexOf("if (s.paused) {", autoSrc.indexOf("// If resuming from paused state"));
 assertTrue(resumeSectionStart > 0, "auto.ts has the paused-session resume block");
 
-const resumeSectionEnd = autoSrc.indexOf("await autoLoop(", resumeSectionStart);
-assertTrue(resumeSectionEnd > resumeSectionStart, "resume block reaches autoLoop");
+const resumeSectionEndCandidates = [
+  autoSrc.indexOf("await runAutoLoopWithUok(", resumeSectionStart),
+  autoSrc.indexOf("await autoLoop(", resumeSectionStart),
+].filter((idx) => idx > resumeSectionStart);
+const resumeSectionEnd = resumeSectionEndCandidates.length > 0 ? Math.min(...resumeSectionEndCandidates) : -1;
+assertTrue(resumeSectionEnd > resumeSectionStart, "resume block reaches the dispatch loop");
 
 const resumeSection = autoSrc.slice(resumeSectionStart, resumeSectionEnd);
 

package/src/resources/extensions/gsd/tests/commands-backlog.test.ts

@@ -0,0 +1,158 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+import { mkdirSync, writeFileSync, readFileSync, existsSync, rmSync } from "node:fs";
+import { join } from "node:path";
+import { tmpdir } from "node:os";
+import { randomUUID } from "node:crypto";
+
+// ─── Helpers ──────────────────────────────────────────────────────────────
+
+function makeTmpBase(): string {
+  const base = join(tmpdir(), `gsd-backlog-test-${randomUUID()}`);
+  mkdirSync(join(base, ".gsd"), { recursive: true });
+  return base;
+}
+
+function cleanup(base: string): void {
+  try { rmSync(base, { recursive: true, force: true }); } catch { /* */ }
+}
+
+function backlogPath(base: string): string {
+  return join(base, ".gsd", "BACKLOG.md");
+}
+
+function writeBacklog(base: string, content: string): void {
+  writeFileSync(backlogPath(base), content, "utf-8");
+}
+
+function readBacklog(base: string): string {
+  return readFileSync(backlogPath(base), "utf-8");
+}
+
+// Test the parsing/writing logic inline since the handler requires runtime context
+
+interface BacklogItem {
+  id: string;
+  title: string;
+  done: boolean;
+  note: string;
+}
+
+function parseBacklog(content: string): BacklogItem[] {
+  const items: BacklogItem[] = [];
+  for (const line of content.split("\n")) {
+    const match = line.match(/^- \[([ x])\] (999\.\d+) — (.+?)(?:\s*\((.+)\))?$/);
+    if (match) {
+      items.push({
+        id: match[2],
+        title: match[3].trim(),
+        done: match[1] === "x",
+        note: match[4] ?? "",
+      });
+    }
+  }
+  return items;
+}
+
+function formatBacklog(items: BacklogItem[]): string {
+  const lines = ["# Backlog\n"];
+  for (const item of items) {
+    const check = item.done ? "x" : " ";
+    const note = item.note ? ` (${item.note})` : "";
+    lines.push(`- [${check}] ${item.id} — ${item.title}${note}`);
+  }
+  lines.push("");
+  return lines.join("\n");
+}
+
+// ─── Tests ──────────────────────────────────────────────────────────────
+
+test("backlog: parse empty file returns empty array", () => {
+  const items = parseBacklog("");
+  assert.equal(items.length, 0);
+});
+
+test("backlog: parse valid entries", () => {
+  const content = `# Backlog
+
+- [ ] 999.1 — OAuth support (added 2026-03-23)
+- [x] 999.2 — Rate limiting (promoted 2026-03-24)
+- [ ] 999.3 — Dark mode`;
+
+  const items = parseBacklog(content);
+  assert.equal(items.length, 3);
+  assert.equal(items[0].id, "999.1");
+  assert.equal(items[0].title, "OAuth support");
+  assert.equal(items[0].done, false);
+  assert.equal(items[0].note, "added 2026-03-23");
+
+  assert.equal(items[1].id, "999.2");
+  assert.equal(items[1].done, true);
+  assert.equal(items[1].note, "promoted 2026-03-24");
+
+  assert.equal(items[2].id, "999.3");
+  assert.equal(items[2].title, "Dark mode");
+  assert.equal(items[2].note, "");
+});
+
+test("backlog: format roundtrips correctly", () => {
+  const items: BacklogItem[] = [
+    { id: "999.1", title: "OAuth support", done: false, note: "added 2026-03-23" },
+    { id: "999.2", title: "Rate limiting", done: true, note: "promoted 2026-03-24" },
+  ];
+
+  const formatted = formatBacklog(items);
+  const parsed = parseBacklog(formatted);
+
+  assert.equal(parsed.length, 2);
+  assert.equal(parsed[0].id, "999.1");
+  assert.equal(parsed[0].title, "OAuth support");
+  assert.equal(parsed[1].done, true);
+});
+
+test("backlog: write and read from disk", () => {
+  const base = makeTmpBase();
+  try {
+    const items: BacklogItem[] = [
+      { id: "999.1", title: "Test item", done: false, note: "added 2026-03-23" },
+    ];
+    writeBacklog(base, formatBacklog(items));
+
+    assert.ok(existsSync(backlogPath(base)));
+    const content = readBacklog(base);
+    assert.ok(content.includes("999.1"));
+    assert.ok(content.includes("Test item"));
+  } finally {
+    cleanup(base);
+  }
+});
+
+test("backlog: next ID increments correctly", () => {
+  const items: BacklogItem[] = [
+    { id: "999.1", title: "First", done: false, note: "" },
+    { id: "999.2", title: "Second", done: false, note: "" },
+    { id: "999.5", title: "Fifth", done: false, note: "" },
+  ];
+
+  let maxNum = 0;
+  for (const item of items) {
+    const match = item.id.match(/^999\.(\d+)$/);
+    if (match) {
+      const num = parseInt(match[1], 10);
+      if (num > maxNum) maxNum = num;
+    }
+  }
+  const nextId = `999.${maxNum + 1}`;
+  assert.equal(nextId, "999.6");
+});
+
+test("backlog: empty backlog returns no items", () => {
+  const base = makeTmpBase();
+  try {
+    // No BACKLOG.md exists
+    assert.ok(!existsSync(backlogPath(base)));
+    // Would return empty array
+  } finally {
+    cleanup(base);
+  }
+});

package/src/resources/extensions/gsd/tests/commands-do.test.ts

@@ -0,0 +1,127 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+
+// ─── Mock dispatcher to capture routed commands ─────────────────────────
+
+let lastRouted: string | null = null;
+let lastQuick: string | null = null;
+
+const mockCtx = {
+  ui: {
+    notify: (_msg: string, _level: string) => {},
+  },
+} as any;
+
+// We test the keyword matching logic directly since the handler imports
+// the dispatcher dynamically (which requires the full extension runtime).
+
+// Inline the route-matching logic from commands-do.ts for unit testing.
+interface Route {
+  keywords: string[];
+  command: string;
+}
+
+const ROUTES: Route[] = [
+  { keywords: ["progress", "status", "dashboard", "how far", "where are we"], command: "status" },
+  { keywords: ["auto", "autonomous", "run all", "keep going", "start auto"], command: "auto" },
+  { keywords: ["stop", "halt", "abort"], command: "stop" },
+  { keywords: ["pause", "break", "take a break"], command: "pause" },
+  { keywords: ["history", "past", "what happened", "previous"], command: "history" },
+  { keywords: ["doctor", "health", "diagnose", "check health"], command: "doctor" },
+  { keywords: ["clean up", "cleanup", "remove old", "prune", "tidy"], command: "cleanup" },
+  { keywords: ["ship", "pull request", "create pr", "open pr", "merge"], command: "ship" },
+  { keywords: ["discuss", "talk about", "architecture", "design"], command: "discuss" },
+  { keywords: ["undo", "revert", "rollback", "take back"], command: "undo" },
+  { keywords: ["skip", "skip task", "skip this"], command: "skip" },
+  { keywords: ["visualize", "viz", "graph", "chart", "show graph"], command: "visualize" },
+  { keywords: ["capture", "note", "idea", "thought", "remember"], command: "capture" },
+  { keywords: ["inspect", "database", "sqlite", "db state"], command: "inspect" },
+  { keywords: ["session report", "session summary", "cost summary", "how much"], command: "session-report" },
+  { keywords: ["backlog", "parking lot", "later", "someday"], command: "backlog" },
+  { keywords: ["add tests", "write tests", "generate tests", "test coverage"], command: "add-tests" },
+  { keywords: ["next", "step", "next step", "what's next"], command: "next" },
+];
+
+interface MatchResult {
+  command: string;
+  remainingArgs: string;
+  score: number;
+}
+
+function matchRoute(input: string): MatchResult | null {
+  const lower = input.toLowerCase();
+  let bestMatch: MatchResult | null = null;
+
+  for (const route of ROUTES) {
+    for (const keyword of route.keywords) {
+      if (lower.includes(keyword)) {
+        const score = keyword.length;
+        if (!bestMatch || score > bestMatch.score) {
+          const idx = lower.indexOf(keyword);
+          const remaining = (input.slice(0, idx) + input.slice(idx + keyword.length)).trim();
+          bestMatch = { command: route.command, remainingArgs: remaining, score };
+        }
+      }
+    }
+  }
+
+  return bestMatch;
+}
+
+// ─── Tests ──────────────────────────────────────────────────────────────
+
+test("/gsd do: routes 'show me progress' to status", () => {
+  const match = matchRoute("show me progress");
+  assert.ok(match);
+  assert.equal(match.command, "status");
+});
+
+test("/gsd do: routes 'run autonomously' to auto", () => {
+  const match = matchRoute("run autonomously");
+  assert.ok(match);
+  assert.equal(match.command, "auto");
+});
+
+test("/gsd do: routes 'clean up old branches' to cleanup", () => {
+  const match = matchRoute("clean up old branches");
+  assert.ok(match);
+  assert.equal(match.command, "cleanup");
+  assert.equal(match.remainingArgs, "old branches");
+});
+
+test("/gsd do: routes 'create pr for milestone' to ship", () => {
+  const match = matchRoute("create pr for milestone");
+  assert.ok(match);
+  assert.equal(match.command, "ship");
+});
+
+test("/gsd do: routes 'add tests for S03' to add-tests", () => {
+  const match = matchRoute("add tests for S03");
+  assert.ok(match);
+  assert.equal(match.command, "add-tests");
+});
+
+test("/gsd do: routes 'what is next' to next", () => {
+  const match = matchRoute("what's next");
+  assert.ok(match);
+  assert.equal(match.command, "next");
+});
+
+test("/gsd do: returns null for unrecognized input", () => {
+  const match = matchRoute("florbinate the gizmo");
+  assert.equal(match, null);
+});
+
+test("/gsd do: prefers longer keyword match", () => {
+  // "check health" (12 chars) should beat "health" (6 chars)
+  const match = matchRoute("check health of the system");
+  assert.ok(match);
+  assert.equal(match.command, "doctor");
+  assert.ok(match.score >= 12);
+});
+
+test("/gsd do: routes 'session report' to session-report", () => {
+  const match = matchRoute("show me the session report");
+  assert.ok(match);
+  assert.equal(match.command, "session-report");
+});